Mechanical Properties of Ewe Vertebral Cancellous Bone Compared With Histomorphometry and High-Resolution Computed Tomography Parameters

Bone Vol. 22, No. 6 June 1998:651– 658

Mechanical Properties of Ewe Vertebral Cancellous Bone Compared With Histomorphometry and High-Resolution Computed Tomography Parameters D. MITTON,1 E. CENDRE,2 J.-P. ROUX,3 M. E. ARLOT,3 G. PEIX,2 C. RUMELHART,1 D. BABOT,2 and P. J. MEUNIER3 1 3

Laboratoire de Me´canique des Solides and 2Laboratoire de Controˆle Non Destructif par Rayonnements Ionisants, INSA, Lyon, France Laboratoire d’Histodynamique Osseuse, INSERM, Lyon, France

vasive nature of the imaging modality and the capacity for three-dimensional imaging at arbitrary orientation make HRCT a promising tool in the quantitative assessment of cancellous architecture. (Bone 22:651–658; 1998) © 1998 by Elsevier Science Inc. All rights reserved.

The goal of the present study was to determine if a highresolution computed tomography (HRCT) system with 150 mm resolution was sufficient to predict mechanical properties in ewe lumbar vertebrae. To answer this question, we used a triangular comparison between: HRCT; biomechanics (compression and shear tests); and histomorphometry, which was the reference method for the measurements of morphometric parameters. Two dissected lumbar vertebrae (L-4 and L-5) from 32 ewes were used. Both compressive and shear properties correlated significantly with amount of bone and structural parameters evaluated by histomorphometry (bone volume/tissue volume, trabecular thickness, trabecular separation), but no significant correlation was found with the trabecular number. With our shear test involving the trabecular architecture itself more significant correlations were found with the node-strut analysis parameters than from the compressive test. Significant correlations were also found between HRCT and histological parameters (bone volume/tissue volume, bone surface/bone volume, trabecular separation, trabecular number, total strut length, number of nodes, and number of termini). Correlations between HRCT structural parameters and mechanical properties on L-4 were of the same magnitude as the correlations between the histomorphometric structural parameters and mechanical results on L-5 but with the remarkable advantage that HRCT is a noninvasive method. In spite of the resolution (150 mm) of our HRCT system, which entailed mainly an enlargement of the thinnest trabeculae or their loss during the segmentation process, we obtained coherent relationships between mechanical and tomographic parameters. The thinnest trabeculae probably had little effect on the mechanical strength. Also, this type of resolution allows us to consider the possibility of perfecting an in vivo HRCT system. However, physical density and bone mineral density correlated much better with strength than either classical histomorphometric or tomographic parameters. The current conclusion is fairly negative with respect to the ability of HRCT to assess mechanical properties nondestructively as compared with dual-energy X-ray absorptiometry. But, the nonin-

Key Words: Ewe; Cancellous bone; Mechanical properties; Highresolution computed tomography; Histomorphometry. Introduction Significant relationships between measures of density and mechanical properties of trabecular bone have been reported.2,5,10,14,16,18,27 Unfortunately, the trabecular bone heterogeneity contributes to significant unexplained variance of the mechanical properties when predicted using scalar measures of mass or density. Bone density alone cannot completely explain the variance of the mechanical property results. Quantification of the cancellous bone architecture appears to be a valid approach for determining the unexplained part of the variance. As a result some investigators have attempted to include measures of architecture to account for increased fragility associated with advancing age.4,8,12,13,21,25,28,29,32 In these studies, node-strut analysis parameters, commonly determined from two-dimensional histologic sections21 and structural indices (orientation, anisotropy, etc.) obtained nondestructively with a microcomputed tomography (micro-CT) system12 were compared with mechanical competence in compression. The architecture or organization of the cancellous tissue within the specimens may contribute to its inherent mechanical properties (compressive strength and Young’s modulus).11,12,28,29 To date, to our knowledge, only one study has been done to estimate the dependence of shear properties (in torsion) on structural parameters.9 In that study, the investigators paid attention only to the trabecular orientation, but did not quantify other architectural parameters. Some studies have shown that CT with a resolution close to 50 mm17 or 100 mm7 gave information comparable to that of histological slice preparation. It was of interest to confirm this point with the use of our high-resolution computed tomography (HRCT) system (150 mm). This method enabled the possibility of perfecting an in vivo HRCT system with an acceptable radiation dose for the patient. The goal of the present study was to determine if an HRCT system with 150 mm resolution was sufficient to predict mechanical

Address for correspondence and reprints: Dr. David Mitton, Laboratoire de Me´caniques des Solides, INSA de Lyon, bat 304, 20 av. A. Einstein, 69621 Villeurbanne Cedex, France. E-mail: [email protected] Presented in part at the Seventh International Congress of Bone Morphometry, October 6 –10, 1996, Alghero, Sardinia, Italy. © 1998 by Elsevier Science Inc. All rights reserved.

651

8756-3282/98/$19.00 PII S8756-3282(98)00036-2

652

D. Mitton et al. Biomechanics, HR computed tomography, histomorphometry

Bone Vol. 22, No. 6 June 1998:651– 658

nial and caudal specimens were processed for mechanical tests (L 5 9 mm for compression test, L 5 5 mm for shear test) and the central one (L 5 5 mm) was used for histomorphometry and HRCT. Due to the small size of some vertebrae, 28 specimens were extracted successfully for histomorphometric and HRCT analysis, 26 for the compression tests and 24 for the shear tests. The specimens for the comparison between the tomography and the bone histomorphometry were embedded in methylmethacrylate. A tomographic image (330 mm thick) was performed on a frontal plane. Then, three 7-mm-thick histologic sections, spaced 150 mm apart, were cut from the same zone. Protocol 2 (L-4 Vertebra) This complementary protocol was designed to complete the comparison between mechanical and tomographic analyses. Consequently, only compression tests were performed. The cores of L-4 were cut, using an Isomet low-speed saw, in two cancellous cubes (a 5 7 mm) in the cranial part to perform a frontal tomographic slice (n 5 32) and in the caudal part to perform a mechanical test (n 5 31). One specimen was lost due to machining difficulties. Mechanical Testing

Figure 1. Preparation of the specimens according to the different protocols. (a) Protocol 1. (b) Protocol 2 (HRCT: high-resolution computed tomography).

properties in ewe lumbar vertebrae. To answer this question, we used a triangular comparison between: HRCT; biomechanics (compression and shear tests); and histomorphometry, which was the reference method for the measurement of morphometric parameters. Materials and Methods Sample Preparation Two lumbar vertebrae (L-4 and L-5) were removed at necropsy from 32 ewes (Limousine and Romanov breeds) at the end of their period of reproduction (mean age 8.9 6 0.9 years, range 5.6–10.3; weight 44–74 kg), and stored at 220°C before processing; this does not induce any mechanical changes.20 The ewes, provided by Institut National de Recherche Agronomique, were bred and slaughtered according to recommendations of the European Economic Community. Vertical cores were obtained by drilling the central portion of the frozen lumbar vertebrae (L-5: f 5 7.5 mm, n 5 28; L-4: f 5 10 mm, n 5 32) (Figure 1). Two different protocols were performed. In protocol 1, the histological slices showed some peripheral fragments on the sides of the cylinders due to the drilling of the specimens. These fragments were easily eliminated from the regions of interest of the histological slices, but they were not precisely identified by HRCT analysis because of a too-low resolution, and were counted as “real” bone. So, in protocol 2, to obtain fragment-free specimens the cylinders were cut in cubic specimens with a low-speed diamond blade saw. Protocol 1 (L-5 Vertebra) The frozen cores of L-5 were sectioned in three contiguous cancellous specimens using an Isomet low-speed saw. The cra-

A universal screw-driven machine (Schenck RSA-250) was used for the test. Samples were placed in a saline bath regulated at 37°C and tested along the craniocaudal axis. Tests were run at 0.5 mm/min to minimize viscous effects. As a first step, we examined L-5 vertebrae (protocol 1). The load applied to the samples was measured with a 2000 N load cell (TME, F 501TC) and the corresponding displacement was measured with a linear variable displacement transducer (LVDT) (Deltalab). To avoid flexion phenomenon, the bottom plate was free to rotate. The measurement of the displacement was then corrected with the machine stiffness according to the method proposed by Anderson et al.1 The uniaxial compression test described earlier24 was performed on the cranial sample and enabled the maximum compressive strength (smax) and Young’s modulus (E) to be obtained. The method used for the shear test based on the use of a sharpened tube was developed in a previous article.24 The parameters calculated were the maximum shear strength (tmax) and the mean elementary shear strength (telem). Next, we studied the L-4 vertebrae (protocol 2). To improve the compression test, ten preloading cycles were applied to samples to reach a steady state19 before the destructive test. The reproducibility of the conditioned tests was generally better than that of nonconditioned tests.19 The load cell was replaced with a 5000 N one (TME, F-501TC) to record the maximum load in all cases. Also, we perfected a new immersed extensometer, which was placed directly at the end of the specimen, thus avoiding correction due to the stiffness of the machine. This new extensometer was validated with the conventional method using nondestructive compressive testing on six specimens. As with the L-5 vertebrae the same compressive parameters were determined. Histomorphometry The undecalcified biopsies were processed according to previously published methods.22 Three 7-mm-thick sections were prepared for staining using Goldner’s method. The results were expressed according to the standardized ASBMR nomenclature without correction for obliquity.30 Measurements were performed with an automatic image analyzer (Visiolab 1000, Biocom, France), operating under Microsoft WINDOWS, and equipped with a 3-CCD color camera. The

Bone Vol. 22, No. 6 June 1998:651– 658

D. Mitton et al. Biomechanics, HR computed tomography, histomorphometry

653

Table 1. Histomorphometric, tomographic, and mechanical characteristics on L-5 vertebrae and comparison between histomorphometry and highresolution computed tomography (HRCT) Histomorphometry Parameters Amount of bone Bone volume/TV (%) Structural Trabecular thickness (mm) Trabecular separation (mm) Trabecular number (/mm) Node-strut analysis Total strut length/TV (mm/mm2) Number of node/TV (/mm2) Number of termini/TV (/mm2) Node to node length/TV (mm/mm2) Node to terminus length/TV (mm/mm2) Terminus to terminus length/TV (mm/mm2)

Mean

SD

Median

HRCT Range

Mean

SD

Median

Range

26.12

8.86

24.64

13.8–56.33

37.31b

6.12

36.30

26.01–51.71

180.4 534.8 1.44

46.8 136.0 0.24

170.4 534.4 1.44

106.8–324.2 251.2–847.6 1–2.08

292.2b 507.3 1.28b

22.9 118.2 0.19

295.1 501.1 1.26

250.4–338.6 294.2–773.2 0.96–1.64

1.62 0.93 0.93 0.63 0.31 0.07

0.35 0.54 0.34 0.38 0.14 0.06

1.58 0.84 0.88 0.6 0.29 0.06

1.23b 0.64a 1.33b X X X

0.20 0.29 0.26 X X X

1.19 0.58 1.33 X X X

0.97–2.51 0.28–2.66 0.30–1.55 0.13–1.81 0.08–0.68 0–0.18

0.87–1.67 0.27–1.40 0.84–1.92 X X X

Biomechanics Maximal compressive strength (MPa) Young’s modulus (MPa) Maximal shear strength (MPa) Mean elementary shear strength (MPa)

23.4 1347 5.3 10.3

8.6 576 3.4 6.7

21.1 1271 4.0 8.6

11.5–47 601–2722 1.3–17.1 3.4–33.5

KEY: TV, tissue volume; X, no measurement. Sample: n 5 28, except for compressive tests (n 5 26), and shear tests (n 5 24). Comparison between histomorphometric and tomographic values: ap , 0.008; bp , 0.0005 by Wilcoxon signed-rank test.

detection of bone was performed using prememorized thresholds. Bone volume/tissue volume or cancellous bone volume (BV/TV) represents the percentage of the tissue occupied by bone. Trabecular thickness (Tb.Th), number (Tb.N), and separation (Tb.Sp) were calculated from BV/TV and the perimeter of trabeculae (BS/TV) according to Parfitt’s formulas.31 The values of histological parameters (BV/TV and BS/TV) used in the present study correspond to the mean value, computed from the three histological sections. All the steps of skeletonization and detection of nodes, termini, and associated struts were automatic with interactive correction if necessary. The following parameters were determined: total strut length (TSL) expressed by tissue volume (TV); number of nodes (N.Nd) and of termini (N.Tm) expressed by

TV; and length of different struts (Nd.Nd, Nd.Tm, Tm.Tm) expressed by TV.23 Throughout this article, BV/TV is defined as an “amount of bone parameter” and the term “structural parameters” is used for Tb.Th, Tb.Sp, and Tb.N. In contrast, TSL, N.Nd, N.Tm, Nd.Nd, Nd.Tm, and Tm.Tm, expressed by TV, are called “node-strut analysis parameters.” High-Resolution Computed Tomography This two-dimensional system has been previously described,15 and is based on the principle of fan beam X-ray CT scanners. The object is mounted on a turntable. An X-ray tube (Pantak, 100 kV) in fine focus configuration (0.4 3 0.4 mm2) is used as a source.

Table 2. Pearson product-moment correlations between the transformed mechanical and histomorphometric parameters in L-5 vertebrae (protocol 1) Compression test Variables Amount of bone Bone volume/TV (%) Structural Trabecular thickness (mm) Trabecular separation (mm) Trabecular number (/mm) Node-strut analysis Total strut length/TV (mm/mm2) Number of node/TV (/mm2) Number of termini/TV (/mm2) Node to node length/TV (mm/mm2) Node to terminus length/TV (mm/mm2) Terminus to terminus length/TV (mm/mm2) KEY: TV, tissue volume. a p , 0.008, bp , 0.0005.

Max. compressive strength (MPa)

Young’s modulus (MPa)

Shear test Max. shear strength (MPa)

Mean elementary shear strength (MPa)

r 5 0.72b

r 5 0.64b

r 5 0.80b

r 5 0.74b

r 5 0.82b r 5 20.45 r 5 0.14

r 5 0.59b r 5 20.52a r 5 0.32

r 5 0.75b r 5 20.65b r 5 0.42

r 5 0.66b r 5 20.67b r 5 0.44

r 5 0.28 r 5 0.31 r 5 20.46 r 5 0.34 r 5 20.19 r 5 20.47

r 5 0.41 r 5 0.38 r 5 20.13 r 5 0.38 r 5 0.22 r 5 20.32

r 5 0.49a r 5 0.52a r 5 20.34 r 5 0.50a r 5 20.13 r 5 20.48

r 5 0.52a r 5 0.57a r 5 20.31 r 5 0.56a r 5 20.28 r 5 20.51a

654

D. Mitton et al. Biomechanics, HR computed tomography, histomorphometry

Figure 2. Comparison between a tomographic slice (upper left) and a histological slice (upper right) and the corresponding segmented images (lower left and right, respectively) across a sample taken from the middle portion of a ewe’s L-5 vertebra and included in methylmethacrylate.

The Thomson detector (TH1482) consists of a linear array of 1024 sensitive elements. In each sensitive element, the scintillator converts X-ray photons into visible light, which is detected by photodiodes. The sampling step (225 mm) corresponds to the size of each diode. The width of the line (500 mm) is set by a lead collimator. The distance from the object to the detector is set to one third of the focus-to-detector distance, providing a final magnification of 31.5. Thanks to this enlargement, the spatial resolution at the level of the object is 150 mm along the detector line and in the perpendicular direction, and the thickness of the imaged slice is about 330 mm (to be compared with the corresponding values of 225 and 500 mm at the entrance of the detector). A higher magnification will still improve the resolution but would expand the geometrical unsharpness at the entrance of the detector. Our choice represents a compromise; that is, for a magnification of 31.5, the geometric unsharpness is equal to the pixel size at the entrance of the detector. Segmentation allowed the trabeculae to be separated from the rest of the image. We perfected a method using an edge enhancement process. A high-pass filter was applied to each tomographic slice, which resulted in enhancing both the edges of trabeculae and the noise. The filtered image was then globally thresholded. In this binary image, the trabeculae thickness was close to the one observed on the tomographic gray-scale image, but the noise was high. In parallel, the initial tomographic slice was globally thresholded. The segmented image obtained included an enlargement of trabeculae, but no noise. The final image was obtained by effecting a logical “and” operation between these two binary images. Although the general pattern seemed rather insensitive to changes in the thresholding values, the thickness and continuity of trabeculae did seem affected. A compromise was realized when choosing the thresholding values: A sufficiently high level of threshold was needed to keep the trabeculae of higher gray level (thinner or less dense trabeculae); however, we took care that this level was not too high so as not to enlarge exaggeratedly the trabeculae of lower gray level (thicker or more dense trabeculae). Because of the limitations of the sampling and thresholding procedures, there was the possibility that disconnected trabeculae and single pixels might be present in the image. Their effect on

Bone Vol. 22, No. 6 June 1998:651– 658

correlations between BV/TV, Tb.N, and 2Euler/volume was tested on three-dimensional 50 mm resolution micro-CT images.12 After removing single voxels and disconnected trabeculae, it was found that the magnitude of the correlations between 2Euler/volume, Tb.N, and BV/TV remained the same. We did not try to quantify the effect of these defaults on HRCT parameters; rather, to compensate, we partially corrected them using binary mathematical morphology: a “hit or miss” transformation enabled us to eliminate the isolated pixels, and most of the trabeculae were reconnected by performing a “closing” operation. The comparison between tomography and bone histomorphometry was made by means of classical parameters like BV/TV, Tb.Th, Tb.Sp, Tb.N, TSL, N.Nd, and N.Tm. Using the VISILOG software (distributed by Noesis Co.) BV/TV and BS/TV were measured on segmented tomographic slices. The measured BV/TV was calculated as the number of pixels corresponding to bone divided by the pixel number of the region of interest (ROI). The trabeculae perimeter was measured directly using the VISILOG software, as well as the node-strut analysis parameters, determined on the images of skeleton. Statistics Analysis was performed using UNISTAT 3.0 software on a PC. Data for some variables were not normally distributed, so a logarithmic transformation was applied to all variables. The linear regressions were used to model the relationships between the transformed variables. Because of the interdependence of the histomorphometric or tomographic parameters, multiple regressions could not be used to evaluate mechanical properties. The Wilcoxon signed-rank test was used for the comparison between histomorphometric and HRCT values. Bonferroni’s procedure was used for increasing the degree of significance required for each test carried out, according to the total number of tests performed to obtain an acceptable global a risk. In our study, p , 0.008 was required for statistical significance. This is a very conservative approach. Results The descriptive statistics of the three methods for protocol 1 are presented in Table 1. Mechanical Parameters vs. Histomorphometry (Protocol 1) The correlation coefficients between mechanical and histomorphometric values are detailed in Table 2. Log(smax), log(E), log(tmax), and log(telem) were significantly correlated with the histomorphometric amount of bone and structural parameters, except log(Tb.N). Contrary to compressive strength, the shear properties [log(tmax), log(telem)] were significantly correlated with nodestrut analysis parameters [log(TSL), log(N.Nd), log(Nd.Nd/TV)]. HRCT vs. Histomorphometry (Protocol 1) Figure 2 provides a comparison between tomographic and histological slices and the corresponding segmented images across a sample taken in the middle part of a ewe’s L-5 vertebra and included in methylmethacrylate. We notice a proper resemblance between the two initial images: trabeculae appear enlarged on the tomographic slice, but even the smallest ones are imaged. In contrast, we observe, on the HRCT segmented image, that some

Bone Vol. 22, No. 6 June 1998:651– 658

D. Mitton et al. Biomechanics, HR computed tomography, histomorphometry

655

Figure 3. Pearson product-moment correlations between the transformed histomorphometric and tomographic parameters in L-5 vertebrae.

of the thinnest trabeculae were lost during the segmentation process. The L-5 vertebrae specimens included in this study ranged in bone volume fraction from 13.8% to 56.33% for bone histomorphometry, and from 26.01% to 51.71% for the HRCT technique (Table 1). Trabecular thickness varied from 106.8 to 324.2 mm (histomorphometry), and from 250.4 to 338.6 mm (HRCT). The mean trabecular thickness was obviously different between histomorphometry and HRCT, as was trabecular bone volume. This result is explained by the difference between HRCT and histological slice thicknesses, of 330 and 7 mm, respectively, and by the resolution of the HRCT system. Due to the resolution of 150 mm, which is close to the trabecular thickness, either the thinnest trabeculae appeared enlarged or they were lost during the segmentation process. We observed that the highest values of BV/TV and Tb.Th measured by HRCT and histomorphometry

were similar, whereas the weakest values were overestimated by HRCT. The loss of the thinnest trabeculae caused Tb.N and N.Nd to be underestimated, and N.Tm overestimated by HRCT. Nodestrut analysis on two-dimensional sections may underestimate nodes and create artificial termini,3 and the loss of trabeculae during the tomographic slice segmentation emphasized this effect. Significant correlations were found between most of HRCT and histological parameters, except for Tb.Th values (Figure 3). Mechanical Parameters vs. HRCT (Protocols 1 and 2) In protocol 1, significant correlations were only found between Tb.Th and smax (r 5 0.57, p , 0.008). No significant correlations were found with the other tomographic parameters. For example, log(smax) was not correlated with log(BV/TV), which is, however, a commonly accepted correlation.

656

D. Mitton et al. Biomechanics, HR computed tomography, histomorphometry

Table 3. Pearson product-moment correlations between the transformed mechanical and tomographic parameters in L-4 vertebrae (protocol 2) Compression test

Variables Amount of bone Bone volume/TV (%) Structural Trabecular thickness (mm) Trabecular separation (mm) Trabecular number (/mm) Node-strut analysis Total strut length/TV (mm/mm2) Number of node/TV (/mm2) Number of termini/TV (/mm2)

Max. compressive Young’s modulus strength (MPa) (MPa) r 5 0.68b

r 5 0.45b

r 5 0.72b r 5 20.60b r 5 0.14

r 5 0.46b r 5 20.41 r 5 0.11

r 5 0.29 r 5 0.29 r 5 20.56b

r 5 0.21 r 5 0.27 r 5 20.42

KEY: TV, tissue volume. p , 0.008, bp , 0.0005.

a

On the contrary, in protocol 2, log(smax) and log(E) were significantly correlated with the tomographic amount of bone and structural parameters, except Tb.N. Additionally, log(N.Tm) was the only node-strut analysis parameter evaluated by HRCT that presented significant correlations with log(smax) (r 5 20.56, p , 0.0005), but fell short of significance with log(E) (r 5 20.42, p 5 0.010) (Table 3). The correlations between HRCT structural parameters and mechanical properties of L-4 are of the same magnitude as the correlations between histomorphometric structural parameters and mechanical results of L-5, but with the remarkable advantage that HRCT is a noninvasive method. Discussion This article has investigated the relationships between the classical parameters measured by histomorphometry and HRCT and the compressive and shear properties. HRCT was compared with two destructive techniques. Histomorphometry is usually performed on specimens. Moreover, from a biomechanical point of view, the use of specimens for this study was justified by the

Bone Vol. 22, No. 6 June 1998:651– 658

work of Mosekilde et al.26 They showed that the maximum compressive strength determined on cylindrical cancellous bone specimens (5 mm in height and 7 mm in diameter) excised from the central part of the vertebral bodies and the maximum compressive strength calculated on the adjacent vertebral body were significantly correlated. Consequently, we used cancellous bone samples. In a first protocol, mechanical properties (compressive and shear) were compared with histomorphometric values. Both compressive and shear properties were correlated with histomorphometric amount of bone and structural parameters [log(BV/ TV), log(Tb.Th), log(Tb.Sp)], except for log(smax) and log (Tb.Sp); however, no significant correlation was found with log(Tb.N). The correlation between log(smax) and log(N.Tm) fell short of significance (r 5 20.46, p 5 0.009), and the others were not significant for the compression test. However, shear strength was correlated significantly with log(TSL), log(N.Nd), and log(Nd.Nd/TV). In addition to the advantages of our shear technique (little damage to the specimen and possibility to test very brittle specimens),24 these correlations showed that the shear test involving the cancellous structure itself is more sensitive than the compression test to node-strut analysis parameters. Correlations between compression values and both BV/TV and Tb.Th agreed well with previous studies on human vertebral specimens.6,33 Dempster et al.6 also determined the Tb.Sp and did not find a significant correlation with compression strength. Concerning the node-strut analysis parameters, the correlations shown on ewe vertebral cancellous bone contrasted to those obtained by Mellish et al.21 on human vertebral cancellous bone. For example, N.Nd correlated significantly with mechanical parameters such as smax for human but not for ewe specimens. Besides the different protocols used, these results suggest that there are important architectural and mechanical differences between ewe and human vertebral cancellous bone. For instance, the difference between the N.Nd of 0.93 in our study and 0.49 in another work21 was not of the same magnitude as the difference in smax between the two species (23.4 MPa for the ewe and 3.2 MPa for humans33); therefore, different parameters may influence the mechanical competence of ewe cancellous bone. The second goal of our study was to evaluate the HRCT by referring to the standard histomorphometric technique. The results obtained enabled us to derive coherent relationships be-

Figure 4. Lateral and vertical sampling in tomographic slices.

Bone Vol. 22, No. 6 June 1998:651– 658

tween bone histomorphometry and high-resolution X-ray tomography for most of the parameters. However, there was no significant correlation for the trabecular thickness. The resolution of our tomograph is similar to the mean trabeculae thickness (around 180 mm). The detector is made up of discrete sensitive elements and, due to the lateral sampling, the thinnest trabeculae, in the section plane, were imaged at best with a thickness equal to the pixel size of 150 mm (Figure 4a). Moreover, with regard to histological slices, the thickness of the tomographic slice was about 40 times greater than the histological one. Due to vertical sampling or tomographic slice thickness, trabeculae positioned slantwise in the section thickness appeared enlarged (Figure 4b). This degradation added to the process of lateral sampling. When looking at the mean values of tomographic and histological parameters, there were also differences between our two methods. Indeed, on the tomographic slice, trabeculae were enlarged, so BV/TV and T.Th were overvalued. Moreover, some trabeculae were lost after tomographic slice segmentation, so Tb.N and N.Nd were undervalued. These data again underline the importance of the thresholding procedure. Resolution and thresholding processes showed that parameter changes from one specimen to another were weaker with HRCT than with the histomorphometry technique. BV/TV and the estimation of Tb.Th were more greatly affected by resolution and threshold than a length measurement such as TSL. If very precise results are needed, only the highest resolution will predict the correct values. Ru¨egsegger et al.,34 working on three-dimensional tomographic images, have shown that the strong dependency of the structural properties on the resolution was rather monotonous, and that there was a good chance to correct the values for a specific resolution up to a voxel size of approximately 200 mm. Similar effects were described for twodimensional structural properties.7 In our study, a resolution of 150 mm was chosen to demonstrate the feasibility of an in vivo HRCT system: a better resolution would deliver an unacceptable dose for the patient. According to Engelke et al.,7 the radiation dose remains acceptable for a spatial resolution close to 200 mm. The matching of HRCT slices to histological sections met with practical limitations, due to the HRCT technique, both in the ability to discern the exact position of the reference level, and for location of fragments present in the specimens. Kuhn et al.17 evaluated a 50 mm resolution micro-CT system to assess trabecular bone structure. In their study, the percent differences in quantitative measures from matched micro-CT slices and histological sections were calculated and used as indicators of accuracy. The percent difference was defined as: % difference 5 (tomographic slice measure 2 average histological section measures)/average of the two measures 3 100. We obtained results close to theirs in terms of % difference, except for BV/TV. The percent differences for their micro-CT system and our HRCT system were 1.35% and 38%, respectively, for BV/TV, 15% and 3.3% for Tb.Th, 13.7% and 20.67% for Tb.Sp, and 214.3% and 211.7% for Tb.N. Our study has provided estimates of the present accuracy of our HRCT system. This system includes the whole process of obtaining two-dimensional images of trabecular bone specimens—from acquiring the images to the thresholding of the reconstructed images, and to the measurement of parameters. The aim of the last part of this study was to compare the mechanical properties with the tomographic parameters. Very few significant correlations were found for the L-5 vertebrae. For example, log(smax) was not correlated with log(BV/TV), which is, however, a commonly accepted correlation. In contrast, the correlations between HRCT structural parameters and mechanical properties of L-4 were of the same magnitude as the corre-

D. Mitton et al. Biomechanics, HR computed tomography, histomorphometry

657

lations between histomorphometric structural parameters and mechanical results of L-5, but with the remarkable advantage that HRCT is a noninvasive method. So the protocol used on L-5 could be criticized. Consequently a different protocol of the machining of the specimens was applied to the L-4 vertebrae of the same ewes. Two different specimen locations were conserved as in the protocol 1. Compressive properties and parameters determined by HRCT were calculated on these specimens. This study recommended the use of cubic specimens (avoiding bone fragments), particularly for performing unbiased HRCT analysis. In conclusion, we were aware that only a high resolution (,50 mm) would predict correct values of classical parameters determined by histomorphometry. However, a resolution of 150 mm enabled relationships between mechanical and tomographic parameters to be obtained. Moreover, the resolution of our HRCT system distorted mainly the values of the thin trabeculae, and most were enlarged. The thinnest were even lost but they probably had little effect on mechanical strength. This type of resolution enabled us to consider the possibility of perfecting an in vivo HRCT system. Finally, in a previous study24 on ewe cancellous bone, a very close correlation was found between density (apparent or ash and bone mineral density) and bone strength (r 5 0.93 and r 5 0.86, respectively). Physical density and bone mineral density correlated much better with strength than either classical histomorphometric or tomographic parameters. The current conclusion is fairly negative with respect to the ability of HRCT to assess mechanical properties nondestructively compared with dual-energy X-ray absorptiometry. However, the noninvasive nature of the imaging modality and the capacity for three-dimensional imaging at an arbitrary orientation make HRCT a promising tool in the quantitative assessment of cancellous architecture of bone.

Acknowledgments: This work was supported in part by a grant from Conseil Re´gional Rhoˆne-Alpes (CRT No. H098730000). The authors thank the Groupe Le´pine, Thomson, INRETS, for help with various aspects of data collection and Franc¸oise Peyrin (Laboratoire CREATIS, INSA, Lyon, France) for having supplied the two-dimensional filtered backprojection algorithm for tomographic slice reconstruction.

References 1. Anderson, M. J., Keyak, J. H., and Skinner, H. B. Compressive mechanical properties of human cancellous bone after gamma irradiation. J Bone Jt Surg 74:747–742; 1992. 2. Carter, D. R. and Hayes, W. C. The compressive behaviour of bone as a two-phase porous structure. J Bone Jt Surg 59:954 –962; 1977. 3. Compston, J. E. Connectivity of cancellous bone: Assessment and mechanical implications. Bone 15:463– 466; 1994. 4. Dalstra, M., Huiskes, R., Odgaard, A., and Van Erning, L. Mechanical and structural properties of pelvic trabecular bone. J Biomech 26:523–535; 1993. 5. Deligianni, D. D., Maris, A., and Missirlis, Y. F. Stress relaxation behaviour of trabecular bone specimens. J Biomech 27:1469 –1476; 1994. 6. Dempster, D. W., Ferguson-Pell, M. W., Mellish, R. W. E., Cochran, G. V. B., Xie, F., Fey, C., Horbert, W., Parisien, M., and Lindsay, R. Relationship between bone structure in the iliac crest and bone structure and strength in the lumbar spine. Osteopor Int 3:90 –96; 1993. 7. Engelke, K., Graeff, W., Meiss, L., Hahn, M., and Delling, G. High spatial resolution imaging of bone mineral using computed tomography. Comparison with microradiography and undecalcified histologic section. Invest Radiol 38:341–349; 1993. 8. Feldkamp, L. A., Goldstein, S. A., Parfitt, A. M., Jesion, G., and Kleerekoper, M. The direct examination of three-dimensional bone architecture in vitro by computed tomography. J Bone Miner Res 4:3–11; 1989. 9. Ford, C. M. and Keaveny, T. M. The dependence of shear failure properties of trabecular bone on apparent density and trabecular orientation. J Biomech 29:1309 –1317; 1996.

658

D. Mitton et al. Biomechanics, HR computed tomography, histomorphometry

10. Galante, J., Rostoker, W., and Ray, R. D. Physical properties of trabecular bone. Calcif Tissue Res 5:236 –246; 1970. 11. Goldstein, S. A., Goulet, R., and McCubbrey, D. Measurement and significance of three-dimensional architecture to mechanical integrity of trabecular bone. Calcif Tissue Int 53:127–133; 1993. 12. Goulet, R. W., Goldstein, S. A., Ciarelli, M. J., Kuhn, J. L., Brown, M. B., and Feldkamp, L. A. The relationship between the structural and orthogonal compressive properties of trabecular bone. J Biomech 27:375–389; 1994. 13. Hodgskinson, R. and Currey, J. D. Effects of structural variations on Young’s modulus of non-human cancellous bone. Eng Med 204:43–52; 1990. 14. Hodgskinson, R. and Currey, J. D. Young’s modulus, density and material properties in cancellous bone over a large density range. J Mater Sci Mater Med 3:377–381; 1992. 15. Kaftandjian, V., Peix, G., Babot, D., and Peyrin, F. High resolution X-ray computed tomography using a solid-state linear detector. J X-Ray Sci Technol 6:94 –106; 1996. 16. Keller, T. S. Predicting the compressive mechanical behavior of bone. J Biomech 27:1159 –1168; 1994. 17. Kuhn, J. L., Goldstein, S. A., Feldkamp, L. A., Goulet, R. W., and Jesion, G. Evaluation of a microcomputed tomography system to study trabecular bone structure. J Orthop Res 8:833– 842; 1990. 18. Linde, F., Hvid, I., and Madsen, F. The effect of specimen geometry on the mechanical behaviour of trabecular bone specimens. J Biomech 25:359 –368; 1992. 19. Linde, F., Pongsoipetch, B., Frich, L. H., and Hvid, I. Three-axial strain controlled testing applied to bone specimens from the proximal tibial epiphysis. J Biomech 23:1167–1172; 1990. 20. Linde, F. and Sorensen, C. F. The effect of different storage on the mechanical properties of bone. J Biomech 26:1249 –1252; 1993. 21. Mellish, R. W. E., Ferguson-Pell, M. W., Cochran, G. V. B., Lindsay, R., and Dempster, D. W. A new manual method for assessing two-dimensional cancellous bone structure: Comparison between iliac crest and lumbar vertebra. J Bone Miner Res 6:689 – 696; 1991. 22. Meunier, P. J. 1983 Histomorphometry of the skeleton. In: Peck, W. A., Ed. Bone and Mineral Research. Annual A. A Yearly Survey of Developments in the Field of Bone and Mineral Metabolism. Amsterdam: Excerpta Medica; 1983; 191–222. 23. Meunier, P. J. Assessement of bone turnover by histomorphometry in osteoporosis. In: Riggs, B. L. and Melton, L. J., III, Eds. Osteoporosis: Etiology, Diagnosis and Management. New York: Raven; 1988; 259 –310. 24. Mitton, D., Rumelhart, C., Hans, D., and Meunier, P. J. The effects of density

Bone Vol. 22, No. 6 June 1998:651– 658

25. 26.

27.

28.

29.

30.

31.

32. 33.

34.

and test conditions on measured compression and shear strength of cancellous bone from the lumbar vertebrae of ewes. Med Eng Phys 19:464 – 474; 1997. Mosekilde, Li. Vertebral structure and strength in vivo and in vitro. Calcif Tissue Int 53:121–126; 1993. Mosekilde, Li., and Mosekilde, Le. Normal vertebral body size and compressive strength: Relations to age and to vertebral and iliac trabecular bone compressive strength. Bone 7:207–212; 1986. Mosekilde, Li., Mosekilde, Le., and Danielsen, C. C. Biomechanical competence of vertebral trabecular bone in relation to ash density and age in normal individuals. Bone 8:79 – 85; 1987. Mu¨ller, R. and Ruegsegger, P. Three-dimensional finite element modelling of non-invasively assessed trabecular bone structures. Med Eng Phys 17:126 – 133; 1995. Nakabayashi, Y., Wevers, H. W., Cooke, T. D. V., and Griffin, M. Bone strength and histomorphometry of the distal femur. J Arthroplasty 9:307–315; 1994. Parfitt, A. M., Drezner, M. K., Glorieux, F. H., Kanis, J. A., Malluche, H., Meunier, P. J., Ott, S. M., and Recker, R. R. Bone histomorphometry: Standardization of nomenclature, symbols and units. Report of the ASBMR Histomorphometry Nomenclature Committee. J Bone Miner Res 2:595– 610; 1987. Parfitt, A. M., Mathews, C. H. E., Villanueva, A. R., Kleerekoper, M., Frame, B., and Rao, D. S. Relationships between surface, volume, and thickness of iliac trabecular bone in aging and in osteoroposis: Implications for the microanatomic and cellular mechanisms of bone loss. Am Soc Clin Invest 72:1396 – 1409; 1983. Pugh, J. W., Rose, R. M., and Radin, E. L. Elastic and viscoelastic properties of trabecular bone: Dependence on structure. J Biomech 6:475– 485; 1973. Rho, J. Y., Zerwehk, J. E., and Ashman, R. B. Examination of several techniques for predicting trabecular elastic modulus and ultimate strength in the human lumbar spine. Clin Biomech 9:67–71; 1994. Ru¨egsegger, P., Koller, B., and Mu¨ller, R. A microtomographic system for the nondestructive evaluation of bone architecture. Calcif Tissue Int 58:24 –29; 1996.

Date Received: July 28, 1997 Date Revised: January 29, 1998 Date Accepted: January 30, 1998